AIR Net advances highly autonomous networks with full-stack Agentic AI

ZTE’s AIR Net leverages full-stack Agentic AI, digital twins and large models to accelerate the evolution toward L4 autonomous networks and value-driven operations

Partner Content In today's rapidly evolving digital economy, communication networks have moved far beyond their traditional role as mere "connectivity pipelines", becoming the intelligent bedrock of digital transformation across industries. By 2026, global operators stand at a critical inflection point: driven by the large-scale commercial rollout of 5G-Advanced, the deep integration of large AI models, and the increasing maturity of digital twin technology, autonomous networks are advancing from automation-assisted operations to L4 stage of autonomous decision-making This shift signifies not only a technological leap, but a fundamental reshaping of industry business models.

Zheng Peng, Vice President of ZTE, systematically outlined ZTE's strategic framework and practical roadmap for progressing toward L4 autonomous networks.

Zheng Peng, Vice President of ZTE - Click to enlarge

Q1: The global autonomous network landscape is rapidly advancing toward L4. How do you define the core characteristics and mission of L4 autonomous networks today?

Mr. Zheng emphasized that the core mission of L4 autonomous networks is evolving from enhancing operational efficiency to driving business revenue growth. In 2025, through joint efforts across the industry, substantial progress was achieved in standards development, architectural framework design, and the implementation of key L4 scenarios, thereby completing preliminary technical capability validation and a closed-loop transition from L3 to L4 in practice.

Entering 2026, global operators are entering a pivotal phase in the evolution of L4 autonomous networks. With the large-scale commercialization of 5G-Advanced, deep integration of large AI models into communications, and growing maturity of digital twin technology, autonomous networks are moving beyond operational automation. They are now advancing toward an intelligent closed-loop system encompassing real-time perception, proactive analysis, autonomous decision-making and automated execution, progressively becoming the core engine for business revenue growth, service innovation, and green, low carbon development.

However, the traditional model of localized AI deployment faces clear limitations. In lower-level autonomous networks, AI capabilities are primarily confined to the operations and maintenance layer, while network elements lack intrinsic intelligence. This lead to delayed data perception, poor cross-domain coordination, and an inability to support end-to-end autonomous closed-loop operations.

The full-stack AI architecture is the key to overcoming this bottleneck. By vertically integrating intelligence across network elements, platforms, and applications, this architecture enables deep, end-to-end penetration and coordinated synergy of intelligent capabilities throughout the entire network and service processes. It effectively bridges capability gaps, fully meets the requirements of TM Forum's L4 stage, and lays a solid foundation for large-scale deployment and commercialization of highly autonomous networks.

Q2: ZTE proposes that the "full-stack AI architecture" is central to L4 evolution. Could you elaborate on its meaning and innovations?

Mr. Zheng clarified that full-stack AI is more than just deploying isolated AI algorithms within the network. It represents a systemic, end-to-end intelligent transformation spanning network elements, networks, services, and business layers. It integrates and orchestrates capabilities across the entire technology stack to enable deep synergy and coordinated autonomy. Building on the widespread adoption of L3 capabilities, ZTE is actively advancing R&D and pilot projects for L4 autonomous networks. Through the evolution of its full-stack AI architecture and expansion into high-value scenarios, ZTE aims to shift the paradigm from operational efficiency gains to sustainable business revenue growth, paving the way for truly intelligent networks.

In the newly released Autonomous Network White Paper 2026, ZTE unveiled the AIR Net high-level autonomous network solution. Centered on the core principles of value-driven design, intrinsic intelligence, autonomous decision-making, and automated closed-loop execution, AIR Net integrates large models, intelligent agents, and digital twin technologies to deliver three key upgrades:

Architecture Evolution: Building a New Agentic AI Paradigm

Built on a technological foundation of data engines, AI large model engines, and digital twin engines, ZTE has developed a new Agentic AI architecture that deploys intelligent agents across network elements, networks, and services. This enables deep integration of networks and AI, edge autonomy with global coordination, and drives the evolution of autonomous networks and closed-loop business operations.

Capability Deepening: Strengthening the Full-Stack Intelligent Chain for End-to-End Coordination in Complex Scenarios

AIR Net enhances intelligent hardware to enable real-time perception and remote control at the network-element level. Leveraging the Agentic AI architecture and the A2A-T protocol, it strengthens both single-agent closed-loop execution and multi-agent collaboration, enabling cross-domain and cross-layer intelligent orchestration. By deeply integrating large models, knowledge graphs, and digital twin technologies, the solution establishes a full-stack intelligent chain that delivers end-to-end autonomous coordination in complex scenarios such as fault management, network optimization, and user experience assurance.

Value Reconfiguration: Shifting from Operations to Business to Achieve a Closed Loop of Commercial Value

Leveraging full-stack AI capabilities and comprehensive process reengineering, AIR Net transforms the operation and maintenance system from traditional KPIs (Key Performance Indicators) to an integrated value indicator system comprising KBIs (Key Business Indicators), KEIs (Key Effectiveness Indicators), and KCIs (Key Capability Indicators). This shift enables end-to-end ticket elimination, establishes a closed loop for business value realization, and accelerates the monetization of network capabilities in advanced scenarios such as 5G-Advanced and the Industrial Internet.

Q3: What progress has ZTE made in advancing L4 autonomous networks from pilots to large-scale commercial deployment?

Mr. Zheng noted that ZTE is systematically driving L4 autonomous networks from point breakthroughs to scaled commercial deployment, building a closed loop anchored in standards leadership, platform enablement, and ecosystem collaboration.

• Standards Leadership: ZTE actively contributes to the development of autonomous network standards within leading international organizations such as TM Forum, CCSA, 3GPP, and ITU, where it leads several key initiatives. At the same time, ZTE is advancing R&D on the A2A-T protocol, aiming to establish a unified, multi-agent collaborative interaction framework that enhances cross-system and cross-vendor automation and coordination.
• Platform Enablement: Built on the "1+N Model Matrix" and the "Co-Sight 2.0 Agent Factory", AIR Net has developed two types of industrial-grade intelligent agents: Copilots and Experts, enabling L4 high-value pilot applications in service provisioning, change monitoring, energy efficiency optimization, and fault monitoring. These agents deliver network productivity that is scalable, controllable, and value-generating.

• For example, in fault monitoring scenarios, the end-to-end fault handling agents enable cross-domain and cross-vendor fault identification, root cause analysis, and precise demarcation and location across wireless and transmission. This has increased the accuracy of complex fault demarcation and location to 91% and reduced MTTR by 11%, significantly improving operational efficiency and user experience. This capability will be extended to end-to-end fault management across IP, core network, and network cloud.

• Ecosystem Collaboration: ZTE and China Mobile have jointly established the "Innovation+" Autonomous Network Open Lab. By deeply integrating knowledge graphs with strong reasoning models, the lab tackles the highly precise, highly reliable network graph model, significantly enhancing autonomous decision-making capabilities of high-value scenarios. These achievements have been successfully replicated and deployed in regions including Beijing, Guangdong, Jiangsu, Shandong, and Hunan, laying a solid foundation for the large-scale rollout of L4 capabilities.

Moving forward, by leveraging AIR Net, ZTE will fully support continuous evolution and accelerate the large-scale commercial deployment of highly autonomous networks.

Contributed by ZTE.